The evolutionary level of human violence

September 30, 2016 • 9:45 am

There’s a new paper in Nature about the level of intraspecific violence in humans and other species, written by José Maria Gómez et al. (free reference and download below).  The question is how often members of single species kill each other in the wild, and whether humans are outliers. It’s already gotten a lot of attention in the press, including an Atlantic summary by Ed Yong, but I’ve avoided reading the journalism until I read the original paper. Now that I have, I’ll summarize the Nature paper briefly for those who haven’t seen other pieces about it.

First, the authors used data from the literature to estimate the level of lethal violence in 1024 species of mammals from 137 families. The question was this: what percentage of individuals who die within a species do so after interacting with members of their own species? That’s the measure the authors take as the degree of “lethal violence” within species. It does not include lethal violence from members of other species, like rabbits getting nicked by raptors.

When you impose that data on the known phylogeny (family tree) of mammals from genetic and morphological data, you can then, using techniques known for a while, estimate what degree of lethal violence existed in various species’ ancestors.  As a hypothetical example, imagine a group of ten related birds, nine of which have crests and one of which was uncrested. Assume further that we know from genetic data that those birds all had a single common ancestor and were all each other’s closest relatives (i.e. they’re a “monophyletic group”). If that’s the case, then it’s a reasonable assumption that that common ancestor also had a crest. (It’s more parsimonious to assume that the crest was lost once than that it evolved nine or so times independently in the descendants of an uncrested bird.) That’s a simple example, but you can use techniques like that to make quantitative estimates, too, and that’s what the authors did for lethal violence.

They first imposed measured levels of lethal violence on the known phylogeny of mammal species. Here it is; the caption comes from the paper, and the color in a branch indicates the estimated level of violence in that branch, ranging from light yellow (peaceful) to dark red (violent). Click to enlarge, and notice the redness around carnivores and, especially, primates; more on that later:

Tree showing the phylogenetic estimation of the level of lethal aggression in mammals (n = 1,024 species) using stochastic mapping. Lethal aggression increases with the intensity of the colour, from yellow to dark red. Light grey indicates the absence of lethal aggression. Mammalian ancestral nodes compared with human lethal violence are shown in red, whereas main placental lineages are marked with black nodes. The red triangle indicates the phylogenetic position of humans.

The authors also found that related species tended to have related levels of violence. That’s what I would have expected, and when I read that sentence I thought, “Well of course: violence is more common in species that are more territorial as well as those that are more social, for territoriality and sociality breeds inter- and intragroup competition for mates, food, and territory. And of course if a species is social or territorial, its relatives are likely to be social and territorial.”

And, sure enough, there was a strong correlation between both sociality, territoriality, and violence among the species. Here’s a graph showing that, with territoriality seemingly inciting more violence than sociality:

(From paper): The figure shows the phylogenetically corrected level of lethal aggression per group (mean ± s.e.m) and the number of mammalian species included in each group. We used a phylogenetic generalized linear model (PGLS) to test the effect of territoriality (yes or no) and social behaviour (social or solitary) on lethal aggression. The level of lethal aggression was more intense in social and territorial species (PGLS, P < 0.05 in all cases and mammal phylogenies; Extended Data Table 1), with no interaction between these two terms (Extended Data Table 1).

Now what about our own lineage? Information about lethality was obtained from 600 human populations dating from the Paleolithic to the present, using both fossil (bone) and historical evidence. Lethal violence included homicide, cannibalism, war, infanticide, execution, and so on. Information was also available from H. neanderthalensis.  There are two main results:

  • The proportion of individuals in the genus Homo killed by lethal violence was about 2%, and this estimate is robust to things like the uncertainty of phylogenies. This is pretty high compared to some other animals (see below), but is explained by the fact that hominins are both social and territorial. This is the ancestral condition before we became civilized. These levels persist in many non-“civilized” groups, though, and from this the authors conclude that there is an evolved, genetically-based propensity for humans to be violent at a level that causes roughly 1 in 50 humans to be killed violently by other humans. That baseline level can be reduced by the imposition of law and “civilized” societies.
  • Violence is correlated with human social organization. The authors divided human groups into four types: “bands” (hunter-gatherers and the like), “tribes” (small groups that live in semipermanent places, with egalitarian societies composed of hunter/horticulturalists), “chiefdoms” (hierarchical non-industrial societies pervaded by kinship ties), and “states” (“politically organized complex societies”). Here are the data, which show that “historic” bands and tribes didn’t differ significantly from the phylogenetic “ancestral” level of violence, while historic chiefdoms and contemporary bands and tribes have significantly higher levels of violence than presumed in our ancestors. In contrast, both historic and contemporary states have considerably lower levels of violence than the ancestral estimate, probably (as the authors note) because in such societies the state takes over the imposition of violence. That, in fact, is one of Steve Pinker’s hypotheses in Better Angels for the historical decline in violence over the last five centuries.
Human lethal violence in different socio-political organizations28. In all cases the boxplots show median values, 50th percentile values (box outline), 95th percentile values (whiskers), and outlier values (circles). We tested whether the level of lethal violence observed in each ancestral node, human period and human socio-political organization differed significantly from the phylogenetic inferences in a.

Finally, I still haven’t read Ed Yong’s piece, though I will now, but I will reproduce a figure from his piece that someone put on Twi**er. It shows the level in violence among many species, and you’ll be surprised at the most violent:


Yes, the primates are up there, but Jebus, the most violent species is the MEERKAT, with over 19% of individuals killed by other meerkats. Who knew?

Vicious murderers!


Gómez, J. M., M. Verdú, A. González-Megías, and M. Méndez. 2016. The phylogenetic roots of human lethal violence. Nature, doi:10.1038/nature19758, Published online, 28 September 2016.

53 thoughts on “The evolutionary level of human violence

    1. Simples!

      Researchers are obviously publishing in an obscure dead language just like the penguin papers from the turn of the last century

  1. Just guessing but the meerkat is very much a group society. Invasion by another group or some one outside the group means war. Therefore, this close group system creates a lot of violence.

    1. The phylogenetic tree that is shown requires that every branch bifurcates – splits specifically into 2. If you do that, and plot a graph to the depth of 10, you get 2^10 tips (or leaves) to the tree. If you have any other number than 2^n, then the final branching level will be missing leaves, making the overall plot a bit untidy.

      It would (and here we leave my area of expertise in data structures) probably also make the computations from the tree significantly more difficult. Some branches would have a different depth from the others, therefore statistics would be weighted as more significant or less significant to other branches for no reason other than their depth.

    1. Meerkat Manor is a great series about the Kalahari Meerkat Project and I’ve watched all it. It shows how females kill the pups of other females but you never see it actually happening, perhaps because they never caught it on film.

      Meerkats live in what I think of as the worst neighborhood in the world, and they’re probably in the worst possible position in the food chain, being predators and prey at the same time, digging for scorpions and millipedes while watching for attackers on the ground and in the air.

      Life for them is tough and they’re ruthless animals. They’re cute, perhaps partly because they seem to be smiling, but that smile hides a very nasty-looking pair of incisors.

  2. There is a distinction IMO between the quantity of acts of violence and the level of fear they evoke due to how sensational they are.

    For example, air travel is statistically much safer than road travel, but air accidents evoke a level of fear and anxiety that road accidents do not.

    Violent crime in America has definitely decreased, but James Holmes shooting up a movie theatre, and the Boston bomber decrease people’s sense of safety more than a routine bank robbery or burglary.

    1. I aint a-scared of no stinkin’ meerkats.

      Violent crime in America has definitely decreased,

      People’s perception of crime is mostly driven by headlines rather than data. Sigh.

      James Holmes … Boston bomber decrease people’s sense of safety more than a routine bank robbery or burglary.

      Crazy = unpredictable, therefore harder to deal with.

      People also freak out a lot more, er, react more strongly to 10 people getting shot than 10 people killed in a tornado.

  3. I have some concerns about the method they used to infer ancestral states on the phylogeny. In particular, look at the platypus, at 3 o’clock on the figure. There is a long, undivided branch leading to the platypus that starts out fairly lethal, briefly gets less lethal, becomes fairly lethal again, and then finally settles down to less lethal. I can’t think of any valid reason why a method that inferred ancestral states could infer this kind of oscillating change within a single, unbroken, branch. (Caveat– perhaps there are some other branches that can’t be seen in the tiny diagram, but I have a copy of the original paper, and I can’t see any others there, and the platypus branch should be long and unbroken.) Now, this doesn’t discount the general thrust of the paper– certain branches of the mammalian tree are more violent than others, and primates are one of the more violent ones– but it does make me wonder about their more precise, quantitative results.

  4. The meerkat ….., they look so cute.

    I’m very pleased to see that humans only kill other animals on a grand scale; seems that we have a soft spot for other humans.

  5. The ability of a species to apply deadly force, and the vulnerability of conspecifics is also relevant, I expect. Small herbivores like rabbits are not likely very deadly toward each other, even if they are rivals for a head of lettuce.

    1. I was going to raise the same issue. If humans had to kill others with bare hands and teeth, perhaps the mortality rate would be even lower. I expect we are the only mammals that regularly use efficient tools for the task.

      1. Okay. I have another example. Ducks. No teeth and no claws. I bet a duck cannot kill another adult duck. Although they sure do make a lot of noise fighting. I live near a lake. I think they are fighting, but what do I know?

  6. For me, the two bat species really shocked me. Granted, I know next to nothing about bats but that’s something that never would have crossed my mind.

    1. I know North American insectivorous bats fairly well. The males can be total pricks and can be quite territorial, and the males of some species are known to occasionally kill other males. In general, however, bats are relatively gentle creatures. But they are highly social and it makes sense that in species where resource competition is high, violence might be expected.

  7. These results fairly scream out for Weird Al to do a parody of the Captain & Tennille tune and call it “Meerkat Love.” (And speaking of meerkat love, according to the Wikipedia article about them, meerkat breeding foreplay is essentially rape. The females engage in no precopulatory display; males simply fight with them until they submit.)

  8. I had never heard of a diademed sifaka (Propithecus diadema). Turns out it is an endangered lemur and also one of the largest lemur species.

  9. We humans (mostly the males) are nasty pieces of work. Not based on science, in which I am deficient, but on observation and experience, which over 30 years of military and police work provided. We are chimps with more convoluted brains. And yes, I have read Dawkins, and I admire his work. Now if we could only teach our children scientific truths instead of nationalistic delusions, perhaps there would be hope for us.

  10. Hugging is an interesting behavior. I believe that it is the thing that separates us from… the ungulates, for one.

  11. “Incisors don’t kill rabbits, rabbits kill rabbits.”

    From the diagram, there appears to be two lagomorph species that kill their own. As an owner of four different rabbits (domesticated European rabbit species) in the past, I’m not surprised by this.
    This species of rabbit in the wild are both social and territorial, especially the females who will defend their turf against other rabbits, to protect their young and makes sure there is adequate food. The point of a fight between rabbits is that the weaker/lower status rabbit will run off quickly.
    But in an enclosed space like a bedroom, where two domesticated rabbits are meeting for the first time, if they take a dislike to each other, a fight will ensue. I’ve twice seen each of my female rabbits attack a male rabbit, latching onto their necks and biting hard. One female took a chunk of flesh out of the neck of another bunny. To break up the fights, I had to pry the rabbits apart with a broom. I never could let the either female out of her pen unless the other rabbits were safely secured in theirs.
    I’ve also been on the receiving end of a targeted rabbit bite, bestowed by a female rabbit who got angry that I had stopped petting her and gone back to reading my newspaper. I never made that mistake again. Ever.
    Yeah, rabbits look cute. But get them angry – watch out.

  12. Ok so I watch a lot of natural history…
    this was in part about a big grizzly called Van who lived in Alaska, ruler of the beach where salmon just happen to use to get up river to their spawning grounds. Many bears started to arrive and all waiting for the spawning run which may have had something to do with what happened below.

    A female of his choosing played coy with him for weeks and lead him around teasing him as it were. Along came another female, fatal mistake. For some reason this big male chased her down and dispatched her to animal heaven in a violent display of power, guess what happened then? the coy female was immediately available for mating.
    Make what you will of that encounter.

  13. Back in the late 50’s and early 60’s, “science said” that all aggression, including lethal aggression, was purely learned behavior and/or an artifact of culture. For example, according to a paper by John Paul Scott published by the University of Chicago Press, “All research findings point to the fact that there is no physiological evidence of any internal need or spontaneous driving force for fighting; that all stimulation for aggression eventually comes from the forces present in the external environment.” Ashley Montagu added the following “scientific fact” about chimpanzees in his “Man and Aggression,” published in 1968: “The field studies of Schaller on the gorilla, of Goodall on the chimpanzee, of Harrison on the orang-utan, as well as those of others, show these creatures to be anything but irascible. All the field observers agree that these creatures are amiable and quite unaggressive, and there is not the least reason to suppose that man’s pre-human primate ancestors were in any way different.”

    A number of people writing around the same time begged to differ. Strangely enough, no one seems able to recall their names today.

  14. >In contrast, both historic and contemporary states have considerably lower levels of violence than the ancestral estimate, probably (as the authors note) because in such societies the state takes over the imposition of violence. That, in fact, is one of Steve Pinker’s hypotheses in Better Angels for the historical decline in violence over the last five centuries.

    There are few errors in this paragraph. The fact that you shortened Steven Pinker’s own and his book’s name was probably a deliberate choice of style.

    But the hypothesis isn’t Pinker’s, and he doesn’t even claim it to be. But very openly he puts it forth as Hobbes’ Leviathan hypothesis.

    And although Pinker brings forth the argument and studies supporting the Leviathan hypothesis, he after concludes, that the state seems to be only a correlation. The actual causality is between free markets and non-violence. You can simply interpret it in both ways because the two are so strongly correlated. But you can remove the state and still keep the correlation, not the other way around.

    Sorry for nitpicking, but I’m just such a fan of both Pinker and this book.

      1. If you really are curious and don’t have access to the book yourself, I can go and find the citations at some point. But that will be a rather troublesome favour that I’m nevertheless willing to perform for a fellow scholar who’s curious.

        But, then again, if it was a rhetorical question or a gesture of you wishing to argue about the subject matter with me, maybe because your political leanings felt a sting, for instance, then I’m perfectly happy to leave it as that.

        So how is it, friend?

  15. Thanks for the summary! It’s interesting that you and Ed Yong made a lot of similar points, but also had some different areas of focus too. I think I learned something from both articles.

Leave a Reply